{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:B7EA2KR2COOVIQSTJPFLIGST5B","short_pith_number":"pith:B7EA2KR2","canonical_record":{"source":{"id":"2410.10343","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-14T09:58:29Z","cross_cats_sorted":[],"title_canon_sha256":"9e73b9b814e7504181321f140478d144a354064c76b477fa3f4000f0547ba480","abstract_canon_sha256":"adf1b382340413bca1138f4be972b7a57785e29e0c90a3caffc4bf73f8637f17"},"schema_version":"1.0"},"canonical_sha256":"0fc80d2a3a139d5442534bcab41a53e85e15b6b6777c3c2d97ee3780d270a8c1","source":{"kind":"arxiv","id":"2410.10343","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.10343","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"arxiv_version","alias_value":"2410.10343v1","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.10343","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"pith_short_12","alias_value":"B7EA2KR2COOV","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"pith_short_16","alias_value":"B7EA2KR2COOVIQST","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"pith_short_8","alias_value":"B7EA2KR2","created_at":"2026-07-05T09:20:15Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:B7EA2KR2COOVIQSTJPFLIGST5B","target":"record","payload":{"canonical_record":{"source":{"id":"2410.10343","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-14T09:58:29Z","cross_cats_sorted":[],"title_canon_sha256":"9e73b9b814e7504181321f140478d144a354064c76b477fa3f4000f0547ba480","abstract_canon_sha256":"adf1b382340413bca1138f4be972b7a57785e29e0c90a3caffc4bf73f8637f17"},"schema_version":"1.0"},"canonical_sha256":"0fc80d2a3a139d5442534bcab41a53e85e15b6b6777c3c2d97ee3780d270a8c1","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:20:15.082593Z","signature_b64":"07PSzmEDTMF+ml2rPl+7IP8g5eJrw5qCfEonk7hYTIo3/5BDmZNaXWE7p7bV5FR8wORBs/L/Guun9X4h3I9tCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0fc80d2a3a139d5442534bcab41a53e85e15b6b6777c3c2d97ee3780d270a8c1","last_reissued_at":"2026-07-05T09:20:15.082220Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:20:15.082220Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2410.10343","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T09:20:15Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"u+yDDvZI/Z9iLFZOrvP0o0pB/BhZ1pRRFs40xn8JxDDBQs0Qz7G5IzXTjKRZxCFi5V5d21bKcpnpgAA5ksJuCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T12:30:30.977435Z"},"content_sha256":"9234340357510f3284d0eb45e5744d9073eb884d7c8e837379a662e4b69010a1","schema_version":"1.0","event_id":"sha256:9234340357510f3284d0eb45e5744d9073eb884d7c8e837379a662e4b69010a1"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:B7EA2KR2COOVIQSTJPFLIGST5B","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Locking Down the Finetuned LLMs Safety","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Linyi Yang, Minjun Zhu, Ningyu Zhang, Yifan Wei, Yue Zhang","submitted_at":"2024-10-14T09:58:29Z","abstract_excerpt":"Fine-tuning large language models (LLMs) on additional datasets is often necessary to optimize them for specific downstream tasks. However, existing safety alignment measures, which restrict harmful behavior during inference, are insufficient to mitigate safety risks during fine-tuning. Alarmingly, fine-tuning with just 10 toxic sentences can make models comply with harmful instructions. We introduce SafetyLock, a novel alignment intervention method that maintains robust safety post-fine-tuning through efficient and transferable mechanisms. SafetyLock leverages our discovery that fine-tuned mo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.10343","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2410.10343/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T09:20:15Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tmUUOHj1TNyC1PhvNOGNee3e4lfC191S4l8JvfeywHKd2m4EbMJ/xx2qNx778jgGptw68I5QZhANY8RLYKrECw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T12:30:30.977953Z"},"content_sha256":"3f332564c8bf2b198613193ccb79b98dd456c0fe1cc1831dba76a3db5d2a6fb9","schema_version":"1.0","event_id":"sha256:3f332564c8bf2b198613193ccb79b98dd456c0fe1cc1831dba76a3db5d2a6fb9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/B7EA2KR2COOVIQSTJPFLIGST5B/bundle.json","state_url":"https://pith.science/pith/B7EA2KR2COOVIQSTJPFLIGST5B/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/B7EA2KR2COOVIQSTJPFLIGST5B/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-10T12:30:30Z","links":{"resolver":"https://pith.science/pith/B7EA2KR2COOVIQSTJPFLIGST5B","bundle":"https://pith.science/pith/B7EA2KR2COOVIQSTJPFLIGST5B/bundle.json","state":"https://pith.science/pith/B7EA2KR2COOVIQSTJPFLIGST5B/state.json","well_known_bundle":"https://pith.science/.well-known/pith/B7EA2KR2COOVIQSTJPFLIGST5B/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:B7EA2KR2COOVIQSTJPFLIGST5B","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"adf1b382340413bca1138f4be972b7a57785e29e0c90a3caffc4bf73f8637f17","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-14T09:58:29Z","title_canon_sha256":"9e73b9b814e7504181321f140478d144a354064c76b477fa3f4000f0547ba480"},"schema_version":"1.0","source":{"id":"2410.10343","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.10343","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"arxiv_version","alias_value":"2410.10343v1","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.10343","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"pith_short_12","alias_value":"B7EA2KR2COOV","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"pith_short_16","alias_value":"B7EA2KR2COOVIQST","created_at":"2026-07-05T09:20:15Z"},{"alias_kind":"pith_short_8","alias_value":"B7EA2KR2","created_at":"2026-07-05T09:20:15Z"}],"graph_snapshots":[{"event_id":"sha256:3f332564c8bf2b198613193ccb79b98dd456c0fe1cc1831dba76a3db5d2a6fb9","target":"graph","created_at":"2026-07-05T09:20:15Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2410.10343/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Fine-tuning large language models (LLMs) on additional datasets is often necessary to optimize them for specific downstream tasks. However, existing safety alignment measures, which restrict harmful behavior during inference, are insufficient to mitigate safety risks during fine-tuning. Alarmingly, fine-tuning with just 10 toxic sentences can make models comply with harmful instructions. We introduce SafetyLock, a novel alignment intervention method that maintains robust safety post-fine-tuning through efficient and transferable mechanisms. SafetyLock leverages our discovery that fine-tuned mo","authors_text":"Linyi Yang, Minjun Zhu, Ningyu Zhang, Yifan Wei, Yue Zhang","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-14T09:58:29Z","title":"Locking Down the Finetuned LLMs Safety"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.10343","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:9234340357510f3284d0eb45e5744d9073eb884d7c8e837379a662e4b69010a1","target":"record","created_at":"2026-07-05T09:20:15Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"adf1b382340413bca1138f4be972b7a57785e29e0c90a3caffc4bf73f8637f17","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-14T09:58:29Z","title_canon_sha256":"9e73b9b814e7504181321f140478d144a354064c76b477fa3f4000f0547ba480"},"schema_version":"1.0","source":{"id":"2410.10343","kind":"arxiv","version":1}},"canonical_sha256":"0fc80d2a3a139d5442534bcab41a53e85e15b6b6777c3c2d97ee3780d270a8c1","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"0fc80d2a3a139d5442534bcab41a53e85e15b6b6777c3c2d97ee3780d270a8c1","first_computed_at":"2026-07-05T09:20:15.082220Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:20:15.082220Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"07PSzmEDTMF+ml2rPl+7IP8g5eJrw5qCfEonk7hYTIo3/5BDmZNaXWE7p7bV5FR8wORBs/L/Guun9X4h3I9tCg==","signature_status":"signed_v1","signed_at":"2026-07-05T09:20:15.082593Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.10343","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:9234340357510f3284d0eb45e5744d9073eb884d7c8e837379a662e4b69010a1","sha256:3f332564c8bf2b198613193ccb79b98dd456c0fe1cc1831dba76a3db5d2a6fb9"],"state_sha256":"4d7f0c195456d88bb6b305cb8342113d85edaa96c5e8360cdbdbf1dd5e2dfa9e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"brf9JqVuSN6PKcf5Eao7y5ld9CMukg+JLdVt+g6Pz4W4WNkSR3WyHfZ9tkJZnk4RPvUVHmJfZ3VcxBM8JH0yDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T12:30:30.984543Z","bundle_sha256":"b0cf7fed130cd113767310b8726c3eaf82871fb5d62a9c86e3f537f89225165b"}}